Evaluating Accuracy and Reliability of Brain-Behavior Models Using Diffusion MRI

McKenzie Paige Hagen1, , John Kruper1, Keshav Motwani2, Eardi Lila2, Jason Yeatman3, Ariel Rokem1


1 Department of Psychology, University of Washington
2 Department of Biostatistics, University of Washington
3 Graduate School of Education, Stanford University

Background

  • Diffusion MRI (dMRI) measures tissue properties of white matter, which contains long-range connections between different brain regions.
  • Brain-behavior models can be used link neuroimaging features and phenotypes.

Question: How do sets of features derived from different dMRI processing methods compare in model accuracy and variability?

Methods

  • Diffusion MRI from 1041 Human Connectome Project participants
  • Processed into “tract profiles” using pyAFQ (cite) and “local connectome” features using DSI-Studio (cite).
  • LASSO models were run on both tract profiles and local connectome.
  • Sparse Group LASSO (SGL) models run on only tract profiles.
  • Prediction targets were various cognitive phenotypes.
  • Models implemented using R and trained using nested cross-validation and boostrap resampling.
Caption for the picture.
Caption for the picture.

Conclusions (rough wording)

  • The selection of model and feature set might not be influential on the accuracy, but may result in less variable, more interpretable models.
  • Tract profiles and local connectome have similar accuracies, but the grouping of tract profiles combined with SGL is good*
  • Splitting families across the train/test splits is bad practice, but didn’t have a large effect on the outcome.

Caption for the picture. Figure 1. This figure needs a title. This is an informative caption, which tells you things about the figure.

Caption for the picture. Figure 2. This figure needs a title. This is an informative caption, which tells you things about the figure.

Acknowledgements

  • Data were provided by the Human Connectome Project, WU-Minn Consortium

  • Krell logo, NIH logo

References

Created with (Allaire et al. 2024)

Allaire, JJ, Yihui Xie, Christophe Dervieux, Jonathan McPherson, Javier Luraschi, Kevin Ushey, Aron Atkins, et al. 2024. Rmarkdown: Dynamic Documents for r. https://github.com/rstudio/rmarkdown.